[OpenCV] Samples 03: kmeans

注意Mat作为kmeans的参数的含义。

扩展：高维向量的聚类。

一、像素聚类

#include "opencv2/highgui.hpp"
#include "opencv2/core.hpp"
#include "opencv2/imgproc.hpp"
#include <iostream>

using namespace cv;
using namespace std;

// static void help()
// {
//     cout << "\nThis program demonstrates kmeans clustering.\n"
//             "It generates an image with random points, then assigns a random number of cluster\n"
//             "centers and uses kmeans to move those cluster centers to their representitive location\n"
//             "Call\n"
//             "./kmeans\n" << endl;
// }

int main( int /*argc*/, char** /*argv*/ )
{
    const int MAX_CLUSTERS = 5;
    Scalar colorTab[] =
    {
        Scalar(0, 0, 255),
        Scalar(0,255,0),
        Scalar(255,100,100),
        Scalar(255,0,255),
        Scalar(0,255,255)
    };

    Mat img(500, 500, CV_8UC3);
    RNG rng(12345);

    for(;;)
    {
        //Jeff --> The second parameter is non-inclusive boundary.
        int k, clusterCount = rng.uniform(2, MAX_CLUSTERS+1);
        int i, sampleCount = rng.uniform(2, 1001);
//        int i, sampleCount = 10;

        Mat points(sampleCount, 1, CV_32FC2), labels;

　　　　 //一般来说，没有必要。sampleCount都远大于ClusterCount。
//        clusterCount = MIN(clusterCount, sampleCount);
        Mat centers;

        /* Jeff --> generate random sample from multigaussian distribution 以某一个中心点，二维高斯分布分配点；主要是一个数学技巧。*/
        for( k = 0; k < clusterCount; k++ )
        {
            Point center;
            center.x = rng.uniform(0, img.cols);
            center.y = rng.uniform(0, img.rows);

            Mat pointChunk = points.rowRange(k*sampleCount/clusterCount,
                                             k == clusterCount - 1 ? sampleCount :
                                             (k+1)*sampleCount/clusterCount);
            rng.fill(pointChunk, RNG::NORMAL, Scalar(center.x, center.y), Scalar(img.cols*0.05, img.rows*0.05));
            cout << pointChunk << endl;
        }

        //洗牌
        randShuffle(points, 1, &rng);

        std::cout << points << std::endl;

        //Jeff --> Mat is vector here, including a list of points.
        //         labels: index of cluster for each points.
        kmeans(points, clusterCount, labels,
            TermCriteria( TermCriteria::EPS+TermCriteria::COUNT, 10, 1.0),
               3, KMEANS_PP_CENTERS, centers);

        //Jeff --> Draw point (tiny circle) with its color on black background.
        img = Scalar::all(0);

        // Step One: show sample points.
        for( i = 0; i < sampleCount; i++ )
        {
            int clusterIdx = labels.at<int>(i);
            Point ipt = points.at<Point2f>(i);
            circle( img, ipt, 2, colorTab[clusterIdx], FILLED, LINE_AA );
        }

        // Step Two: show central points.
        for( i = 0; i < clusterCount; i++ )
        {
            std::cout << centers.at<Point2f>(i) << std::endl;
        }

        imshow("clusters", img);

        char key = (char)waitKey();
        if( key == 27 || key == 'q' || key == 'Q' ) // 'ESC'
            break;
    }

    return 0;
}

二、图像的kmeans降维处理

g++ -std=c++11 -pthread  -fpermissive  main.cpp -o output `pkg-config --cflags --libs opencv` -ldl

From: http://seiya-kumada.blogspot.com/2013/03/k-means-clustering.html【非常好】

#include <opencv2/highgui.hpp>
#include <opencv2/core.hpp>
#include <opencv2/imgproc.hpp>
#include <iostream>

using namespace cv;
using namespace std;

void show_result(const cv::Mat& labels, const cv::Mat& centers, int height, int width)
{
        std::cout << "===\n";
        std::cout << "labels: " << labels.rows << " " << labels.cols << std::endl;
        std::cout << "centers: " << centers.rows << " " << centers.cols << std::endl;
        assert(labels.type() == CV_32SC1);
        assert(centers.type() == CV_32FC1);

        cv::Mat rgb_image(height, width, CV_8UC3);
        cv::MatIterator_<cv::Vec3b> rgb_first = rgb_image.begin<cv::Vec3b>();
        cv::MatIterator_<cv::Vec3b> rgb_last = rgb_image.end<cv::Vec3b>();
        cv::MatConstIterator_<int> label_first = labels.begin<int>();

        cv::Mat centers_u8;
        centers.convertTo(centers_u8, CV_8UC1, 255.0);
        cv::Mat centers_u8c3 = centers_u8.reshape();

        while ( rgb_first != rgb_last ) {
                const cv::Vec3b& rgb = centers_u8c3.ptr<cv::Vec3b>(*label_first)[];
                *rgb_first = rgb;
                ++rgb_first;
                ++label_first;
        }
        cv::imshow("tmp", rgb_image);
        cv::imwrite("./result.jpg", rgb_image);
        cv::waitKey();
}

int main(int argc, const char * argv[])
{
        cv::Mat image = cv::imread("./d1.jpg");
        if ( image.empty() ) {
                std::cout << "unable to load an input image\n";
                return ;
        }

        std::cout << "image: " << image.rows << ", " << image.cols << std::endl;
        assert(image.type() == CV_8UC3);
        cv::imshow("image", image);

        cv::Mat reshaped_image = image.reshape(, image.cols * image.rows);
        std::cout << "reshaped image: " << reshaped_image.rows << ", " << reshaped_image.cols << std::endl;
        assert(reshaped_image.type() == CV_8UC1);
        //check0(image, reshaped_image);

        cv::Mat reshaped_image32f;
        reshaped_image.convertTo(reshaped_image32f, CV_32FC1, 1.0 / 255.0);
        std::cout << "reshaped image 32f: " << reshaped_image32f.rows << ", " << reshaped_image32f.cols << std::endl;
        assert(reshaped_image32f.type() == CV_32FC1);

        cv::Mat labels;
        int cluster_number = ;
        cv::TermCriteria criteria {cv::TermCriteria::COUNT, , };
        cv::Mat centers;
        cv::kmeans(reshaped_image32f, cluster_number, labels, criteria, , cv::KMEANS_RANDOM_CENTERS, centers);

        show_result(labels, centers, image.rows, image.cols);

        return ;
}

三、ROI的kmeans支持

原文：https://blog.csdn.net/fengbingchun/article/details/79395298

double kmeans( InputArray data, int K, InputOutputArray bestLabels, 
                 TermCriteria criteria, 
                 int attempts, int flags, OutputArray centers = noArray() );

接口的声明在include/opencv2/core.hpp文件中，实现在modules/core/src/kmeans.cpp文件中

(1)、data：为cv::Mat类型，每行代表一个样本，即特征，即mat.cols=特征长度，mat.rows=样本数，数据类型仅支持float；

(2)、K：指定聚类时划分为几类；

(3)、bestLabels：为cv::Mat类型，是一个长度为(样本数，1)的矩阵，即mat.cols=1,mat.rows=样本数；为K-Means算法的结果输出，指定每一个样本聚类到哪一个label中；

(4)、criteria：TermCriteria类，算法进行迭代时终止的条件，可以指定最大迭代次数，也可以指定预期的精度，也可以这两种同时指定；

(5)、attempts：指定K-Means算法执行的次数，每次算法执行的结果是不一样的，选择最好的那次结果输出；

(6)、flags：初始化均值点的方法，目前支持三种：KMEANS_RANDOM_CENTERS、KMEANS_PP_CENTERS、KMEANS_USE_INITIAL_LABELS；

(7)、centers：为cv::Mat类型，输出最终的均值点，mat.cols=特征长度，mat.rols=K.

    // Color dimension reduction
    Mat processTagByKmean(Mat3b const tag, Option option)
    {
        int K = option.knnClusterNum;

        // 0. Prepare arguments for kmeans.
        cv::Mat reshaped_tag = tag.reshape(, tag.cols * tag.rows);

        cv::Mat reshaped_tag32f, labels, centers;
        reshaped_tag.convertTo(reshaped_tag32f, CV_32FC1, 1.0 / 255.0);
        // ------------------------------------------------------------------
        // 1. do kmeans
        cv::kmeans(reshaped_tag32f, K, labels,
                   TermCriteria(CV_TERMCRIT_EPS + CV_TERMCRIT_ITER, , 1.0),
                   , KMEANS_PP_CENTERS, centers);
        // ------------------------------------------------------------------
        // 2. convert to rgb mat
        cv::Mat rgb_tag(tag.rows, tag.cols, CV_8UC3);
        cv::MatIterator_<cv::Vec3b> rgb_first  = rgb_tag.begin<cv::Vec3b>();
        cv::MatIterator_<cv::Vec3b> rgb_last   = rgb_tag.end<cv::Vec3b>();
        cv::MatConstIterator_<int> label_first = labels.begin<int>();

        cv::Mat centers_u8;
        centers.convertTo(centers_u8, CV_8UC1, 255.0);
        cv::Mat centers_u8c3 = centers_u8.reshape();

        while (rgb_first != rgb_last)
        {
            const cv::Vec3b &rgb = centers_u8c3.ptr<cv::Vec3b>(*label_first)[];
            *rgb_first = rgb;
            ++rgb_first;
            ++label_first;
        }

        return rgb_tag;
    }

原文：https://blog.csdn.net/qq_22764813/article/details/52135686

如果Mat类型数据的深度和通道数不满足上面的要求，则需要使用convertTo()函数和cvtColor()函数来进行转换。
convertTo()函数负责转换数据类型不同的Mat，即可以将类似float型的Mat转换到imwrite()函数能够接受的类型。
而cvtColor()函数是负责转换不同通道的Mat，因为该函数的第4个参数就可以设置目的Mat数据的通道数（只是我们一般没有用到它，一般情况下这个函数是用来进行色彩空间转换的）。
另外也可以不用imwrite()函数来存图片数据，可以直接用通用的XML IO接口函数将数据存在XML或者YXML中。